ECW output & -co "TARGET="

[bolsen@…]

In the current cvs version specifing a target doesn't really do anything. I tried numbers from 0 to 100 and all seemed to default to highest quality.

I would also suggest that TARGET be specified as a rate instead of a percentage, more consisitent with the ER Mapper ECW documentation. When I hacked out the percentage and directly assigned to rate the compression worked ok.

It might be helpful to throw out some debug as to what the actual compression rate was since their compressor actually gives you back performance feedback values.

ie, nCompressionRate=1 is highest quality, seen anywhere from 2:1 - 5:1 nCompressionRate 15 seen anywhere from 18:1 to 25:1

[warmerdam]

Brian,

Are you using GDAL 1.2.6 or CVS?  There have been some changes to handling
of the TARGET value since 1.2.6 release I think. 

I use the TARGET formulation so it is more like the compression used
for other formats, even though it is different than the normal ECW
representation.  

[bolsen@…]

Using current cvs.  I checked out afternoon 20050706.

Another suggestion might be to allow gdal standardization on "TARGET" specifying
desired output size in terms of % of origina and also provide a
"TARGET_COMPRESSION" or comparable option to specify a "native" number.

All the different ways all of these different formats like to specify
compression is confusing.

[warmerdam]

Brian,

I'll make an effort to keep consistency in use of TARGET, but I'm not keen 
on adding a variety way of setting the same thing. 

Thanks,

[bolsen@…]

I have to reopen this one.  It's causing users way too much confusion.

Based on user's normal experience with MrSID, Global Mapper, Er Mapper, etc, a
user wanted to use gdal to test various compression rates.

Giving a formula to a user probably isn't the best thing in the world:

gdal_target = 100 * (compression rate - 1) / (compression rate)

Using what they're used to corresponds to :

1, 5, 7, 10, 15, 20, 30, 50 (this is compression rate).

the gdal equivalent being:

0, 80, 85.714, 90, 93.333, 96.667, 98 (jpeg style compression)

There just has to be some way to take into account that wavelets are handled in
a fundamentally different way from traditional forms of compression.

[Jukka Rahkonen]

TARGET is documented in ​http://www.gdal.org/frmt_ecw.html and if this ticket has not seen any activity in nine years it can well be closed.